AUTOMATED MALARIA PARASITE DETECTION BASED ON IMAGE PROCESSING PROJECT REFERENCE NO.: 38S1511 COLLEGE : BANGALORE INSTITUTE OF TECHNOLOGY, BENGALURU BRANCH : COMPUTER SCIENCE AND ENGINEERING GUIDE : DR. ASHA. T STUDENTS : MR. DEEPAK R MR. SWAROOPARADHYA M C MS. AISHWARYA M G MS. GEETHA B S Keywords: Image Acquisition,Pre-processing, Feature Extraction, Segmentation,Morphological Operations, cell differentiation Introduction: Malaria is a mosquito borne disease caused by parasites of genus plasmodium. The person gets affected by malaria when malaria parasites are introduced into circulatory system by infected female anopheles mosquito bites. Diagnosis of malaria parasitemia from blood smears is a subjective and time-consuming task for pathologists. The automatic diagnostic process will reduce the diagnostic time, also it can be worked as a second opinion for pathologists and may be useful in malaria screening. This project presents an automatic method for malaria diagnosis from thin blood smears. Loading the image is the first phase,later the image is being preprocessed to remove unwanted noise and brightness. Feature extraction and successive segmentation techniques are then applied on the image to focus on important parts of the image. Finally Morphological operations are carried out to differentiate parasite cells from the RBC cells and their respective count is determined.
Objectives: We aim at Automated malaria parasite detection based on image processing, to introduce fast and accurate method for malaria parasite detection. A single blood smear image can be processed multiple times for various detection of blood components unlike or original blood sample using both thick and thin films, The objective are to differentiate RBC and the infected cells which are present in blood smear side, mitigate problems posed by different conditions such as noisy and degrading to set on efficient and awareness result. Better image enhancement algorithm and Morphological operation are proposed to get the count of RBC and parasite in order to speed up the diagnosis process. Input: Digitalized malaria blood smear image Output: Differentiate normal and infected cell and give the count of total RBC and parasites infected cells. Methodology: The architectural details for the proposed project are explained as shown in Figure 1 Image Acquisition Collect Malaria samples and give as input Preprocessing RGB2GrayS cale Median Filter Histogram Equalization Segmentation Removing Noise Intensity Adjustment Binary Image Morphology Operation Edge Detection Dilate Region Filling Thresholding Finding Perimeter Cover surface of the cells Cell Counting Based on threshold value differentiate normal and infected cells and get the count of cells Figure 1 Architecture diagram
Image Acquisition In any image processing project, data collection plays an important role. Finding the required datasets is a prime task. The data preparation typically consumes about 90% of the time of the project. Once available data sources are identified, they need to be selected, cleaned, constructed and formatted into the desired form. One of the forms of image acquisition in image processing is known as real-time image acquisition. This usually involves capturing real images obtained from blood samples of malaria patients preparing slides of the smear and digitalizing it. This digital image is used as a source. Pre-processing The segment of input image of (250 250) pixels is selected for further processing. The input image may have low brightness and contrast. Real-world images are highly susceptible to noisy, missing, and inconsistent data. Low-quality data will lead to low-quality mining results. Hence it is essential to pre-process the data. There are a number of pre-processing techniques. In our work, we mainly aim at median filter and histogram equalization. Segmentation Segmentation divides the image into its constituent regions or objects. The goal of segmentation is to simplify and/or change the representation of an image into something that is more meaningful and easier to analyze. Image segmentation is typically used to locate objects and boundaries in images. In our study we first remove noise, adjust intensity of the image, perform gray threshold and convert the image to binary form. Morphology Operation Morphological operations are image processing operations which processes images based on shapes. It applies a structuring element of specific shape and size on input image. The output image is created by comparing the value of each pixel with its neighbours. These operations are sensitive to the shape of the structuring. Further operations such as holes filling, overlaying is carried out which helps in detection of infected cells.
Cell Counting By comparing the overlay of original image and masked image and based on the intensity profile, differentiation between the normal and infected cells is carried out. Finally using mathematical operations, the count of number of such cells is determined and is displayed. Result and Conclusion: The experiment was conducted by collecting blood samples of patients suffering from malaria. The microscope connected to the personal computer was used to view the blood films and these films were digitalized. Totally 10 samples are taken from different blood samples. The images were used as raw data for malaria parasite count. The result of some sample images is reported in Table1. Image No Manual RBC Count Table 1 Result Analysis IP Approach RBC Count IP approach count of malaria parasite Manual count of malaria parasite Difference in algorithmic count and manual count (%) Image 1 20 23 4 4 0 Image 2 120 118 32 35 2.54 Image 3 62 59 12 15 5.08 Image 4 286 290 44 41 1.03 Image 5 803 820 137 139 0.24 Image 6 63 67 24 21 4.47 Image 7 70 72 2 3 1.38 Image 8 104 106 34 38 3.77 Image 9 148 153 138 137 0.65 Image 10 98 106 2 1 0.94 From the Performance analysis it is found that the system gains accuracy of 85.54%, sensitivity of 95.5%, specificity of 80.8%, recall 95.5 and F-score of 73.90.
Conclusion: The proposed system provides a robust automated system for detection of malaria parasites in thin and thick blood films. The detection of Malaria parasites is done by pathologists manually using microscopes. So, the chances of false detection due to human error are high, which in turn can result into fatal condition. This system curbs the human error while detecting the presence of malaria parasites in the blood sample by using image processing techniques. We achieved this goal by using Image Segmentation, Morphological operations, edge detection technique to detect malaria parasites in images acquired from digitalized blood samples. The system acts in a robust manner so that it is unaffected by the exceptional conditions and achieved high percentages of sensitivity, specificity and prediction values. After implementation of the proposed approach for the Lab sample images and for the available image database, it is found that the parasites count is near about matching with the manual count while in the RBCs count, some more differences are observed. Future Scope: The system which is at present developed using MATLAB software can further be implemented in android platform. Support Vector Machine (SVM) techniques can be used to analyze and classify the parasite species. The system can be used as a helping guide for physicians in laboratory, especially in places where there are less experts related to the field.